Modulation of Interferon Action by RetinoidsINDUCTION OF MURINE STAT1 GENE EXPRESSION BY RETINOIC ACID*

(Received for publication, August 13, 1996, and in revised form, January 10, 1997)

Xiao Weihua‡, Venkatadri Kolla‡§, and Dhananjaya V. Kalvakolanu¶

From the University of Maryland Cancer Center, Department of Microbiology & Immunology, Program in Oncology,Molecular and Cellular Biology Program, University of Maryland School of Medicine, Baltimore, MD 21201

We have previously demonstrated that up-regulationof STAT1 protein by all-trans-retinoic acid (RA) in inter-feron (IFN)-unresponsive cells permits growth inhibi-tion by IFNs. Here, we show that the promoter of STAT1directly responds to retinoic acid treatment. Sequenceand functional analysis of the murine STAT1 promoterhave identified a direct repeat motif that serves as aretinoic acid response element. Mutagenesis of this ele-ment resulted in a loss of response to RA. This element isactivated by RA receptors a, b, and g. In vivo, RA recep-tor b and retinoid X receptor a preferentially interactedwith this element. Thus, these data define a molecularbasis for the synergy between IFNs and retinoids intumor growth inhibition.

Interferons (IFNs)1 regulate cellular antiviral, antitumoral,and immunological responses via specific gene induction (1).Ligand-bound IFN receptors induce the tyrosine phosphoryla-tion of cytoplasmic STAT (signal transducers and activators oftranscription) proteins employing Janus kinases (JAKs). Acti-vated STATs then migrate to the nucleus and bind to specificresponse elements to induce the expression of IFN-stimulatedgenes (2). IFNs-a/b induce cellular genes via an IFN-stimu-lated response element, which binds several transacting factorsof which ISGF-3 is the primary regulator (2). This factor con-sists of a 48-kDa DNA binding protein (ISGF3g) and threetyrosine phosphorylated proteins, STAT1a, STAT1b, andSTAT2 (2). Two Janus kinases, tyk2 and JAK1, are essentialfor the activation of STATs in response to IFN-a/b. Ligand-stimulated IFN-g receptor, employing Janus kinases JAK1 andJAK2, induces phosphorylation of STAT1 but not STAT2.STAT1 then migrates to nucleus, binds to g-IFN-activated site,and induces gene expression (2). Cell mutants lacking STAT1and JAK1 fail to respond to IFN-a/b and IFN-g (3, 4). Thus, JAK1

and STAT1 are common signaling components for all IFNs.All-trans-retinoic acid (RA), a potent biological response

modifying metabolite of vitamin A, induces cellular gene ex-pression utilizing nuclear receptors that also act as transcrip-tion factors (5). These receptors bind to retinoic acid responseelements (RARE) to activate gene transcription. Retinoic acidreceptors (RAR) and retinoid X receptors (RXR) are the twomajor mediators of retinoid actions (5). Three major forms ofRARs and RXRs, a, b, and g, and the corresponding subtypesare known. Their expression is variable depending on cell type,organ, and status of cell differentiation. The preferred ligandsfor RAR and RXRs are RA and 9-cis-RA respectively, althoughRA at high concentrations activates RXR (5). RXRs form het-eromeric complexes with RARs, bind to RAREs, and stimulatespecific gene transcription (5). Heteromeric complexes of RXRsformed with other nuclear receptors such as those of thyroidhormone (TR), vitamin D3, and peroxisome proliferator activa-tor also induce ligand-specific gene expression (6–10). Thus,RXRs are important cofactors for nuclear receptor-mediatedgene regulation.

Several studies have indicated that IFN and RA in combina-tion produce additive or synergistic antitumoral effects in vivoand in vitro (11). It is not known how these two disparateligands and their corresponding signal transduction systemscross-talk in the mediation of antitumoral activity. We haveshown previously that these effects are in part mediated by anincrease in the level of IFN-stimulated gene factors upon RAtreatment, thus allowing their functional activation by IFN(12). In particular, STAT-1 protein is induced in IFN-resistanttumor cells treated with RA (13). In the present investigation,we have identified the mechanism by which RA up-regulatesthe expression of STAT1.

MATERIALS AND METHODS

Cell Culture and Reagents—All cell lines were cultured in mediasupplemented with charcoal-treated, dialyzed fetal bovine serum. Mu-rine embryonic tumor cell lines F9 and P19 and monkey kidney cell lineCOS-7 were cultured in Dulbecco’s modified Eagle’s medium. HumanRARa, -b, and -g and mouse RXRa expression vectors were provided byPierre Chambon, IGBMC, France. Mouse RXRb expression vector wasdescribed previously (8). Human thyroid hormone receptor beta (TRb)regulated by RSV-LTR was provided by Edwards Park, University ofTennessee. Rabbit anti-STAT1 antibodies were provided by ChrisSchindler, Columbia University. Mouse anti-JAK-1 antibody was fromTransduction Laboratories. Purified baculovirus-expressed RARb andRXRb, mouse monoclonal antibody against RXRb (H2RIIBP), and rab-bit anti-RARb antibody were gifts from Keiko Ozato, National Insti-tutes of Health. Rabbit anti RARa, RARg, and RXRa antibodies werepurchased from Santa Cruz Biotechnology. All these antibodies cross-reacted with cognate proteins from mouse and human sources. MouseIFN-g was from Boehringer Mannheim. Purest preparations of all-trans-retinoic acid (99%), 9-cis-retinoic acid (high performance liquidchromatography pure) and 3,39,5 triiodo-L-thyronine (T3 or thyroid hor-mone) were purchased from Sigma and were reconstituted in ethanol.The oligonucleotides used in this study were shown in Table I. Thesewere used for electrophoretic mobility shift assays (EMSA) and cloning

* This work was supported by a grant from the National CancerInstitute (to D. V. K.). The costs of publication of this article weredefrayed in part by the payment of page charges. This article musttherefore be hereby marked “advertisement” in accordance with 18U.S.C. Section 1734 solely to indicate this fact.

The nucleotide sequence(s) reported in this paper has been submittedto the GenBankTM/EBI Data Bank with accession number(s) U80267.

‡ Contributed equally to this work.§ Present address: Kimmel Cancer Institute, Philadelphia, PA 19107.¶ To whom correspondence should be addressed: E-mail: dkalvako@

umabnet.ab.umd.edu.1 The abbreviations used are: IFN, interferon; DR-5, direct repeat

element with 5 nucleotide spacing; EMSA, electrophoretic mobility shiftassay; JAK, Janus tyrosine kinase; RA, all-trans-retinoic acid; RARE,retinoic acid response element; RAR, retinoic acid receptor; RXR, reti-noid X receptor, STAT, signal transducing activator of transcription;pIRE, palindromic IFN response element; TR, thyroid hormone recep-tor; TK, thymidine kinase; kb, kilobase(s); bp, base pair(s); ISG, IFN-stimulated gene; LTR, long terminal repeat; MHC, major histocompat-ibility complex.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 272, No. 15, Issue of April 11, pp. 9742–9748, 1997© 1997 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www-jbc.stanford.edu/jbc/9742

into TK-luciferase. Probes for EMSA were prepared by filling in theends with Klenow fragment in the presence of [a-32P]dCTP or by endlabeling with [g-32P]ATP. For site-directed mutagenesis, 59-GATAAAT-CAGGGTGATAA-39 (RM1) and 59-TAGAGTTCAGTGTGATAA-39 (RM3)oligos were used.

Gene Expression Analyses—Northern blot, run off transcription, andWestern blot analyses were performed as described elsewhere (12).Cells (2 3 105) were electroporated with 5 mg of luciferase reporterconstruct, 0.5 mg of expression vectors of retinoid receptors, and 2 mg ofb-actin-b-galactosidase plasmid, and luciferase assays were performed(14). b-Galactosidase assays were performed to normalize for variationsin transfection efficiencies. Stable transfection of F9 cells with STAT1cDNA, cloned in mammalian expression pCXN2, was performed asdescribed previously (13). This plasmid also carried a G418 resistancemarker for selection in mammalian cells. The resultant drug-resistantcolonies (;150) were pooled and used in the experiments. COS-7 cellextracts, expressing individual retinoid receptors for EMSA, were pre-pared as described (15).

An adult BALB/c mouse liver genomic library (Clontech) in EMBL3phage vector (5 3 106 plaque-forming units) was screened with a 32P-labeled human STAT1 cDNA (16), and 6 clones were identified. Allthese clones contained the same 18-kb insert as analyzed by restrictiondigestion and Southern blotting. A 4.5-kb XhoI fragment was detectedwhen probed with a 270-bp fragment representing the 59 end of thecDNA. This fragment was cloned into pGL3-basic vector (Promega).pGL3-TK was constructed by excising the TK promoter (109 bp) frompTK-CAT vector and cloning into the pGL3-basic vector. Deletion andpoint mutations were constructed (17) using a polymerase chain reac-tion-based kit (Stratagene) and a reporter construct VKL-7 as template.All constructs were confirmed by sequencing.

RESULTS

Retinoic Acid Modulates the IFN Stimulation of Transcrip-tion Factors in Embryonic Tumor Cells—We have previouslyshown that treatment of F9 embryonal carcinoma cells with RAenhances the ISG transcription (12) due to an activation ofISGF-3 in dF9 (RA-differentiated) but not in F9 (undifferenti-ated) cells. To further understand the basis for the failure ofIFN-induced transcription, we performed EMSAs to detect theactivation of STAT1 by IFN-g in F9 and dF9 cells. As shown inFig. 1, nuclear extracts from IFN-g-treated dF9 but not F9 cells(compare lane 1 with 2) were able to form a complex with a32P-labeled palindromic IFN-g response element (pIRE) (18).Identity of this factor as STAT1 was established by a specificantibody that neutralized the formation of the complex (lane 5).Similar activation of STAT1 was also observed in RA-treatedP19 (dP19) but not in undifferentiated P19 cells (Fig. 1, lanes 3and 4). Mixing of F9 cell extracts with those of dF9 cells did notprevent the binding of STAT1 to pIRE (lane 6).

Since STAT1 was activated by IFN-g after RA-treatment, wewanted to further distinguish whether such activation was due

to an increase in the levels of signal transducing JAKs ortransacting factors. To test these possibilities, F9 cells werestably transfected with pCXN2 expression vector or the samevector that expressed the STAT1 cDNA. The resultant celltransfectants were treated with IFN-g, and their nuclear ex-tracts were assayed for pIRE binding of STAT1 using EMSA(Fig. 2A). Activation of STAT1 was not detected in untreatedcells (lanes 1 and 3). In the F9 transfectants that carried theexpression vector pCXN2 alone (lanes 1 and 2), IFN-g failed toinduce STAT1 binding to pIRE (lane 2). However, IFN-g ro-bustly activated STAT1 in cells transfected with the cDNA(Fig. 2A, lane 4). To further prove that the transfected STAT1was functional, luciferase reporter assays were performed. Asanticipated, no induction of luciferase activity was detected inthe F9 cells that stably expressed the pCXN2 vector alone (Fig.2B, bar 2). In STAT1 transfected cells (bars 3 and 4), IFN-greadily induced the expression of pIRE-luciferase reporter gene(Fig. 2B, bar 4). Thus, F9 cells possessed the necessary recep-tors and protein kinases for the activation of IFN-g-initiatedJAK-STAT pathway, except STAT1. Over expression of STAT1or treatment with RA restored IFN-g responses in these cells.Consistent with this, JAK1 levels were not affected by RA-treatment in an immunoprecipitation assay (Fig. 2C). Westernblot analyses (Fig. 2D) revealed no detectable STAT1 protein inundifferentiated F9 cells. However, it was strongly induced byRA-treatment (lanes 2 and 3). Thus, STAT1 appeared to be atarget of retinoid-mediated regulation.

RA Induces the Expression of STAT1 Gene—Since STAT1protein was induced by RA, we next determined whether suchenhancement was due to an induction of STAT1 gene expres-sion. Northern blot analysis (Fig. 3A) did not reveal detectableSTAT1 mRNA (lanes 1 and 2) in untreated F9 cells or cells thatreceived ethanol (the solvent in which RA was prepared).Treatment of cells with 1 and 10 mM RA strongly induced themRNAs of STAT1a and -b (lanes 3 and 4). The probe detectedboth the mRNAs because they were derived from the same gene(16). All the cells expressed similar levels of b-actin mRNAirrespective of treatments. To examine whether the inductionof STAT1 by RA was due to de novo transcription, nuclear runoff transcription assays (Fig. 3B) were performed. No detecta-

TABLE IOligonucleotide probes used in this study

RW, wild-type RARE of STAT1 promoter. Boldface nucleotides rep-resent the RARE. RM1, RM2, RM3, and RM4 are mutant versions ofRW and where the mutated nucleotides have been italicized; DR-5,consensus RARE (19); H2RII, RARE from mouse major histocompati-bility class I gene (8).

RW 59 TCGAGATGGGTCAGGGTGATAAC 3939 CTACCCAGTCCCACTATTGAGCT 59

RMI 59 TCGAGATAAATCAGGGTGATAAC 3939 CTATTTAGTCCCACTATTGAGCT 59

RM2 59 TCGAGATGGGTCAAAATGATAAC 3939 CTACCCAGTTTTACTATTGAGCT 59

RM3 59 TCGAGATGAGTCAGTGTGATAAC 3939 CTACTCAGTCACACTATTGAGCT 59

RM4 59 GATACCGTCCACATGTAA 3939 CTATGGCAGGTGTACATT 59

H2RII 59 GCCAGGCGGTGAGGTCAGGGGTGGGGAA 3939 TCCGCCACTCCAGTCCCCACCCCTTCGG 59

DR-5 59 GATAGGTCACCAGGAGGTCATAA 3939 CTATCCAGTGGTCCTCCAGTATT 59 FIG. 1. RA enhancement of STAT1 activation by murine IFN-g

in F9 and P19 cells. Cells were first treated with RA (1 mM for 3 days)followed by IFN-g (200 units/ml) for 30 min where indicated. Nuclearextracts (3 mg) were incubated with a 32P-labeled pIRE probe in EMSA.Lane 5 was similar to lane 2 except that the extracts were incubatedwith anti-STAT1 antibody for 40 min at room temperature prior to theaddition of the probe. Lane 6 contained a 1:1 mixture of extracts used inlanes 1 and 2 incubated for 30 min prior to the addition of the probe.Position of the STAT1 complex is indicated. Treated and untreated lanesare indicated as 1 and 2 respectively.

Retinoid Induction of STAT1 Gene 9743

ble transcription of STAT1 was observed in untreated cells.However, RA induced the transcription by 24 h, which in-creased with prolonged treatment (48 h). All these cells ex-pressed normal levels of glyceraldehyde-3-phosphate dehydro-genase transcripts, indicating that lack of STAT1 geneexpression in F9 cells was not due to a global transcriptionalblockade. Further, RA did not induce the expression of STAT2mRNA (Fig. 3C). Induction of STAT1 mRNA by RA was ob-served in an IFN-unresponsive MCF-7 breast carcinoma cellline and an acute promyelocytic leukemia cell line (data notshown). Thus, STAT1 mRNA is up-regulated by RA in multiplecell types.

Identification of a RARE in STAT1 Promoter—To examinewhether the STAT1 promoter was directly regulated by RA, amouse genomic clone was isolated using human STAT1 cDNA(16) as a probe. A 1.85-kb HindIII fragment, containing 700 bpof upstream sequence, the first exon and intron (1.15 kb), wascloned into pGL-3 basic vector. The resultant reporter con-struct, VKL-2, responded to RA in COS-7 cells upon cotrans-fection with RARa (Fig. 4A). Since COS-7 cells lacked thesenuclear receptors, the induction was due to the cotransfected

receptor. Deletion of a 965-bp sequence consisting of first exonand a substantial portion of intron from VKL-2 (constructVKL-4) had no effect on RA inducibility (Fig. 4B). Followingtreatment with RA (1 mM), a strong up-regulation of luciferaseactivity was noted in cells that were cotransfected with RARa.Furthermore, two other nuclear receptors, RXRb and TRb,along with their ligands 9-cis-RA and T3 had no effect on geneexpression (Fig. 4B). Two other constructs, VKL-6 and VKL-7,which contained up to 2950- and 2670-bp upstream elementsof STAT1 promoter, respectively, were also strongly induced byRA in these cells (Fig. 4A). Thus, the cloned fragments con-tained necessary elements for a specific induction of STAT1promoter by RA. Primer extension analyses identified the startsite at 12 nucleotides upstream of ATG codon (data not pre-sented). We then tested the effects of other members of theRAR family on STAT1 promoter. COS-7 cells were transfectedwith RARa, -b, and -g expression vectors along with luciferasereporter VKL-4 (Fig. 4C). Although all the members of the RARfamily induced the reporter gene expression, RARb was aslightly better activator than the others. Luciferase gene didnot express upon transfection of VKL-7 in F9 cells. RA treat-ment for 24 h caused a 12-fold increase in luciferase activity(Fig. 4D). It was further enhanced with longer treatment (48 h).

Sequence analysis of the promoter (Fig. 5) revealed a directrepeat element, at 2467 bp, that could potentially serve asRARE. More significantly, this sequence had closer resem-blance to H2RII of the MHC class I gene (8), than to the otherRAREs. Unlike the previously described retinoid response ele-ments (5), the STAT1-RARE had a near perfect repeat se-quence of GGGTCAGGGTGA with no spacer nucleotides (Fig.5). Further, GGG residues were found in both the half-sites inplace of AGG of most retinoid responsive half-sites. At 2122and 2338 positions, TATA-like elements were present.

Mutational Analysis of the Retinoic Acid Response Element—Since RA stimulated STAT1 and sequence analysis identified aRARE in the promoter, we next examined whether this elementalone was sufficient for RA inducibility. Using a PCR-basedapproach (17), we constructed a deletion mutant that lackedthe direct repeat element in VKL-7 reporter. The wild-typeconstruct (VKL-7) but not the mutant (DM) responded to RA inCOS-7 cells when cotransfected with RARb (Fig. 6A, bars 2–5).

FIG. 2. Modulation of IFN-g sensitivity of F9 cells by RA. A,stable transfection of STAT1 cDNA restored IFN-sensitivity in F9 cells.EMSA was performed as in Fig. 1. Lanes 1 and 2, cells transfected withpCXN2 expression vector alone; lanes 3 and 4, cells that expressedSTAT1 protein. No treatment and IFN-g treatment (200 units/ml for 30min) were indicated with 2 and 1 signs, respectively. Nuclear extracts(3 mg) were used in EMSA. Position of STAT1 complex is indicated. B,functional activity of the transfected STAT1 gene product. Cells used inbars 1–4 were similar to those in lanes 1–4 of panel A. pIRE-luciferease(10 mg) and b-actin-b-galactosidase (2 mg) plasmids were electroporatedinto cells and analyzed for gene expression (13). 30 h after transfection,IFN-g (200 units/ml) treatment was performed for 16 h, where indi-cated, and luciferase activity was measured. Mean 6 S.E. of triplicatemeasurements was presented. Transfection efficiency was normalizedwith b-galactosidase measurements. C, JAK1 levels in F9 cells. Cellswere treated with the indicated concentration of RA for 3 days. Equalamounts of cell extracts (80 mg) were immunoprecipitated using an antiJAK-1 antibody. The proteins were separated on 10% SDS-polyacryl-amide gels and Western blotted. The blots were probed with the sameantibody using the ECL method (Amersham Corp.). D, STAT1 expres-sion in F9 cells. Cell lysates (30 mg) were Western blotted and analyzedwith a rabbit anti-STAT1 antibody.

FIG. 3. Effect of RA on STAT gene expression. A, F9 cells weretreated with the indicated agents for 3 days, and total RNA was iso-lated. An equal amount of RNA (30 mg) was analyzed in Northern blots.The blots were probed with 32P-labeled human STAT1 or b-actin cDNAprobes. Positions of the STAT1a and STAT1b mRNAs are indicated bythe upper and lower arrows, respectively. B, nuclear run-off transcrip-tion. F9 cells were treated with 1 mM RA for the indicated times, andnuclear run-off transcription was performed as described earlier. Totallabeled nuclear RNAs were hybridized for 3 days to Bluescript (negativecontrol), STAT1 and glyceraldehyde-3-phosphate dehydrogenasecDNAs immobilized on nylon membranes. The blots were washed andautoradiographed. C, Northern blot analysis of STAT2 mRNA. TotalRNA (30 mg) from untreated and RA (1 mM for 72 h) treated F9 cellswere Northern blotted and probed with human STAT2 and b-actinprobes.

Retinoid Induction of STAT1 Gene9744

Using the same PCR approach, we generated point mutantswith RM1 and RM2 oligonucleotides in the native promoter(VKL-7). In RM1, GGG residues of the left half-site were re-placed with AAA. RM3 bore mutations in the central G residueof the GGG bases of both the half-sites (see “Materials andMethods”). Transfection of these two point mutants into COS-7cells along with RARb did not significantly induce the promoter(Fig. 6B, bars 3–6). Thus, mutations in the GGG residues ofeither half-site abolished RA inducibility.

We next determined whether fusion of the synthetic STAT1-RARE conferred RA inducibility to a neutral promoter drivingthe expression of luciferase gene. COS-7 cells were transfectedwith the reporter constructs bearing either a wild-type or amutant response element upstream of herpes simplex viral TKgene promoter, along with RARs, and were treated with RA.Insertion of two copies of a wild-type element (RW2X) produceda slightly higher activity than the one with a single copy(RW1X) in COS-7 cells (Fig. 6C, compare bar 4 with 6). Mu-tants that lacked the GGG residues in either half-site wereunresponsive to RA (bars 8, 10, and 12). This observation wasconsistent with the data obtained with the same point mutantsin the context of native promoter (Fig. 6B). Furthermore, mu-tation of a central G residue of both the half-sites (RM3) causeda similar loss of response. Therefore, the direct repeat elementof STAT1 promoter was sufficient for induction by RA.

Binding of Transcriptional Factors to the RA Response Ele-ment—To identify the transcription factors that mediate theseeffects, we performed EMSA. In these experiments, untreatedand RA-treated F9 nuclear extracts were incubated with a32P-labeled synthetic RW (STAT1-RARE) to detect specificDNA binding (6). A slow mobility complex, A, appeared inRA-treated F9 cells (Fig. 7A, lanes 2-5) whose binding wasenhanced with prolonged RA treatment. Binding of this com-

plex to RW was eliminated upon preincubation of these nuclearextracts with the wild type (RW) oligonucleotide (lanes 6–8)but not by a mutant (RM4) oligonucleotide (lanes 9–11). Mu-tant oligos RM1, RM2, and RM3 also failed to compete out thebinding (data not presented). Formation of complex-A was in-hibited by preincubation of extracts with polyclonal antibodiesspecific for RARb and RXRa (lanes 12 and 13) but not by thoseraised against RARa, RARg, and RXRb (lanes 14–16). Bindingof this complex to RW was not eliminated by preincubation ofcell extracts with DR-5 (19) and H2RII oligos at 5 or 25 3concentrations (Fig. 7B, lanes 2, 3, 6, and 7). However, at highconcentrations (100 and 500 3), DR-5 competed with the la-beled RW (lanes 4 and 5), indicating a preferential formation ofcomplex-A with the latter. H2RII failed to compete out the RWcomplex-A (lanes 6–9). RW also formed another complex, B,with untreated F9 cell extracts whose binding disappearedwith RA treatment of cells (Fig. 7A, see lanes 1–3).

RW binding factors in F9 cells exhibited interesting proper-ties compared with those that bound to consensus RARE andDR-5. RW binding of complex-A was RA inducible (Fig. 7C,lanes 2–5), while DR-5 binding factors were constitutive (Fig.7C, lanes 6–9). Formation of complex-A was not detected withuntreated F9 cell extracts (Fig. 7A, lane 1). The DR-5 bindingfactors from untreated F9 cell extracts (lane 6) were recognizedby antibodies against RARa and RARg (data not presented).Under the same conditions, complexes were barely detectablewith H2RII probe (lanes 10–13). Interestingly, complex-B wasformed only with RW upon incubation with untreated F9 ex-tracts but not with DR-5 or H2RII. Binding of this complex waseliminated upon preincubation of cell extracts with RW but notRM3, RM4, DR-5, or H2RII oligos (data not presented). Fur-thermore, complex-B was not recognized by antibodies specificfor RARa, RARb, RARg, RXRa, or RXRb (Fig. 7D). Thus, it

FIG. 4. Analysis of murine STAT1 promoter. A, schematic representation of the STAT1 promoter fragments cloned upstream of luciferasegene in pGL-3 vector. Upstream sequences from the STAT1 gene are indicated with a thick black line. Sequences corresponding to exon and intronare shown with filled boxes. Selected restriction enzyme sites are indicated: H, HindIII; P, PstI. Fold induction of each reporter by RA treatment(1 mM for 16 h) in COS-7 cells are indicated on the right. B, induction of VKL-4 reporter gene by nuclear receptors transfected into COS-7 cells.pSG5 (29) was the control expression vector. RA and 9-cis-RA and T3 were used at 1 mM for 16 h, where indicated. Equal amounts of protein (50mg) from each sample were assayed for luciferase activity. C, induction of VKL-4 reporter in COS-7 cells by the members of the RAR family.Transfection and RA treatment were similar to panel B. D, induction of VKL-7 reporter (15 mg) by RA in transfected F9 cells. RA (1 mM) wasperformed for the indicated time. Cell extracts (80 mg) were assayed for luciferase activity.

Retinoid Induction of STAT1 Gene 9745

appeared to be a novel factor.Cotransfection of RARb and RXRa Induces the RA-inducible

Expression from STAT1 Promoter—Since the above studiesindicated the binding of RARb and RXRa complexes to STAT1-RARE, we next determined whether this combination of recep-tors would augment RA-dependent gene expression from theSTAT1 promoter. VKL-7 reporter vector was cotransfected

with expression vectors for either RARb or RXRa or the com-bination into COS-7 cells. RARb, but not RXRa, alone inducedgene expression (Fig. 8A, bars 2 and 3) upon RA (1 mM) treat-ment. However, the combination of these two receptors causedstronger expression (bar 4). To further test whether such in-duction was a result of binding of these receptors to RW, EMSAwas performed with whole cell extracts of COS-7 cells trans-fected with the individual expression vectors. RARb bound toRW, which was further enhanced when combined with extractsfrom RXRa expressing cells. (See Fig. 8B, lanes 3 and 5). RXRaalone bound very weakly to this element (lane 4). Pre-incuba-tion of these extracts with cognate antibodies eliminated thebinding of these complexes (data not presented). Interestingly,purified RARb (expressed in a baculovirus vector) alone failedto form such a complex with RW, although it bound to DR-5efficiently (data not shown). These data suggest that additionalcellular factors may be necessary for the formation of RWbinding complexes.

DISCUSSION

Dependence of IFN-g responses in F9 cells on RA-treatmentsuggests two possibilities. RA may enrich the levels or activi-ties of IFN-receptor components and the associated Janus ki-nases or of IFN-regulated transcription factors. Since transfec-tion of STAT1 alone restores the IFN-g-inducible geneexpression, modulation of IFN responses by RA may not in-volve an enhancement of JAKs or other receptor components.Consistent with this, RA did not increase IFN receptors densityor avidity in unresponsive cells (20). Furthermore, RA did notinduce the tyrosine phosphorylation of JAK1 (data not shown).Thus, RA modulation of IFN responses occurs at the level ofSTAT1 gene expression in these cells.

Northern and nuclear run off transcription studies haveshown that RA specifically induces transcription of STAT1gene (Fig. 3, A-C). RA does not affect STAT2 expression. Fur-ther, STAT2 promoter does not possess a RARE (21). Analysisof the STAT1 promoter (Figs. 4–6) identified a RARE. Thiselement has several unique properties: i) It is a direct repeatelement with no spacing between the repeats. ii) It is notstimulated by RXRb in the presence of 9-cis-RA or high con-centration of RA that activates RXRs. iii) It preferentially bindsthe RARb and RXRa heteromers, and such binding is increasedwith prolonged treatment of F9 cells with RA. Under the sameconditions RA does not alter the binding of RARzRXR com-plexes to DR-5. iv) In untreated F9 cells, a unique factor (com-plex-B) binds to STAT1-RARE (Fig. 7, A and C) but not to DR-5

FIG. 5. Nucleotide sequence of murine STAT1 gene promoter.Retinoic acid response element is highlighted and underlined. TATA-like elements and ATG codon are shown in italics. Transcriptional startsite is indicated with 11 sign.

FIG. 6. Mutational analysis of the STAT1 promoter. Conditions of transfection analysis were as described under Fig. 4. A, effect of deletionof a direct repeat element. Wt represents VKL-7, and DM represents the same construct with a deletion of the RARE (Fig. 5). B, induction of pointmutants of RARE. Bars 1 and 2, wild-type promoter (VKL-7); bars 3 and 4, mutant bearing the element shown in RM1 oligo; and bars 5 and 6,mutant bearing the element in RM3 oligo. C, synthetic response elements from STAT1 (see Table I) confer RA inducibility to TK promoter-drivenluciferase reporter. Bars 1 and 2, TK-luciferase plasmid; bars 3–6, wild-type element (RW) cloned upstream of TK promoter except that theconstruct used in bars 3 and 4 possessed two tandem copies (RW2X), and the one in bars 5 and 6 contained a single copy (RW1X), respectively; bars7 and 8, RM1 element; bars 9 and 10, RM2 element; and bars 11 and 12, RM3 element. All these constructs contained a single copy of the mutantresponse element. COS-7 cells were transfected and treated as described in Fig. 4.

Retinoid Induction of STAT1 Gene9746

(19) or H2RII (8). The STAT1-RARE requires GGG residues inboth the half sites, since mutagenesis of these in any half siteof the direct repeat element abrogates the RA responses. Itsspecificity for RARb and RXRa is quite intriguing because asimilar oligonucleotide with 5-bp spacing between the directrepeat elements (RW-DR5) has not formed the same complex(data not presented). These data indicate the importance ofclose apposition of the repeat elements for specific binding.Although the direct repeat elements with two or five nucleotidespacing have been shown to respond to RA in several genes (5),the gene for human medium chain acyl coenzyme A dehydro-genase has a unique element with eight nucleotide spacing andanother one with no spacing (22). The human oxytocin andmouse laminin B1 gene promoters also contain RA responsiveelements with 13 and 47 nucleotide spacing (23–25). TheSTAT1-RARE appears to belong to this class of unique nuclearreceptor response elements where spacing between the half-sites is not a primary determinant of retinoid responsiveness(26, 27).

Preferential binding of RARb and RXRa heteromers toSTAT1-RARE may permit a selective regulation of STAT1 geneby RA. This notion is supported by the observation that despite

the abundance of DR-5 binding factors in F9 cells, they did notbind to RW (Fig. 7C). The DR-5 binding factors were recognizedby specific antibodies against RARa and RARg (data not pre-sented). The observations that regulation of transcription bynuclear receptors is dependent on the promoter context, re-sponse element orientation, and DNA binding domain interface(19, 28–35) are also suggestive of such preferential interac-tions. The facts that RARb gene is inducible by RA (36, 37) andthat the binding of RARb to RW is increased upon exposure ofF9 cells to RA also support the notion of preferential binding ofRARbzRXRa to STAT1-RARE. Consistent with these observa-tions, cotransfection and EMSA assays with individual recep-tors in COS-7 cells resulted in a stronger induction of geneexpression (Fig. 8). Since whole cell extracts containing RARb,but not the purified RARb, form specific complexes with RW, itappears that additional cellular factors are necessary for thebinding to occur. Given the unique organization of the repeatelements in STAT1-RARE, such factors may play a crucial rolein the formation of high affinity complexes. Further investiga-tion is necessary to address these issues. Lastly, the F9 cellularfactor (complex-B in Fig. 7A) that binds to STAT1-RARE, maybe a negative regulator since it disappears with RA-treatment

FIG. 7. Electrophoretic mobilityshift assays. A, binding of transactingfactors to wild-type RW (STAT1-RARE) inF9 cells. All treatments are indicated atthe top of the figure. Cell extracts wereprepared after RA treatment (1 mM) forthe indicated times. Nuclear extracts (3mg) from each treatment were incubatedwith 0.25 ng of a 32P-labeled RW (120,000cpm) in EMSA. In lanes 6–8 and 9-11,unlabeled RW and RM4 were used ascompetitors, respectively, prior to the ad-dition of the probe. In lanes 12–18, theextracts were incubated with 2.5 ml ofspecific high affinity antibodies againstretinoid receptors for 40 min at room tem-perature prior to the addition of theprobe. Positions of complexes-A and -Bare indicated on the left. B, specificity oftransacting factor binding to RW. EMSAwas carried out as in panel A except thatindicated amounts of different competitoroligos were preincubated with nuclear ex-tracts of F9 cells treated with RA for 48 h.C, specific activation of binding factors toRW by RA (1 mM) in F9 cells. Nuclearextracts (3 mg) from panel A were incu-bated with different labeled probes indi-cated at the top of the figure. All probeswere labeled to comparable specific activ-ities and used in the experiment. The un-bound DR-5 probe, being the smallest,ran out of the gel in lanes 6–9. D, reactiv-ity of complex-B with antibodies specificto various retinoid receptors. Nuclear ex-tracts of untreated F9 cells (3 mg) wereincubated with the indicated antibodiesas in panel A prior to the addition of theprobe in an EMSA with RW probe. NIS,non-immune serum of rabbit.

Retinoid Induction of STAT1 Gene 9747

and is absent during STAT1 transcription. An important out-come of our study is that it demonstrates for the first time amolecular basis for the modulation of IFN action by RA. Thisobservation indirectly connects the retinoids to cell cycle regu-lation. For example, recent studies indicate that IFNs inhibitcell growth using STAT1 (38) and induce the expression ofp21/WAF/Cip-1 gene, whose product inhibits the cyclin-dependent kinases (39). Elevation of STAT1 levels may thuspermit a robust activation of STAT1 by IFNs, leading to growtharrest.

Acknowledgments—The authors thank all the colleagues who havegenerously provided several valuable reagents used in this study, SaraB. Mannino for technical assistance at the early stages of this work,Rama Kudaravalli for help in transfection experiments, and DanielLindner and Ernest Borden for a critical reading of this manuscript.

REFERENCES

1. Kalvakolanu, D. V., and Borden, E. C. (1996) Cancer Invest. 14, 25–532. Schindler, C., and Darnell, J. E., Jr. (1995) Annu. Rev. Biochem. 64, 621–6513. Muller, M., Briscoe, J., Laxton, C., Guschin, D., Ziemiecki, A., Silvennoinen,

O., Harpur, A. G., Barbieri, G., Witthuhn, B. A., Schindler, C., Pellegrini, S.,Wilks, A. F., Ihle, J. N., Stark, G. R., and Kerr, I. M. (1993) Nature 366,129–135

4. Muller, M., Laxton, C., Briscoe, J., Schindler, C., Improta, T., Darnell, J. E.,Jr., Stark, G. R., and Kerr, I. M. (1993) EMBO J. 12, 4221–4228

5. Mangelsdorf, D. J., and Evans, R. M. (1995) Cell 83, 841–8506. Kliewer, S. A., Umesono, K., Mangelsdorf, D. J., and Evans, R. M. (1992)

Nature 355, 446–449

7. Kliewer, S. A., Umesono, K., Noonan, D. J., Heyman, R. A., and Evans, R. M.(1992) Nature 358, 771–774

8. Marks, M. S., Hallenbeck, P. L., Nagata, T., Segars, J. H., Appella, E.,Nikodem, V. M., and Ozato, K. (1992) EMBO J. 11, 1419–1435

9. Yu, V. C., Delsert, C., Andersen, B., Holloway, J. M., Devary, O. V., Naar, A.M., Kim, S. Y., Boutin, J.-M., Glass, C. K., and Rosenfeld, M. G. (1991) Cell67, 1251–1266

10. Leid, M., Kastner, P., Lyons, R., Nakshatri, H., Saunders, M., Zacharewski, T.,Chen, J.-Y., Staub, A., Garnier, J.-M., Mader, S., and Chambon, P. (1992)Cell 68, 377–395

11. Moore, D. M., Kalvakolanu, D. V., Lippman, S. M., Kavanagh, J. J., Hong, W.K., Borden, E. C., Paredes-Espinoza, M., and Krakoff, I. H. (1994) Semin.Hematol. 31, Suppl. 5, 31–37

12. Kalvakolanu, D. V., and Sen, G. C. (1993) Proc. Natl. Acad. Sci. U. S. A. 90,3167–3171

13. Kolla, V., Lindner, D. J., Weihua, X., Borden, E. C., and Kalvakolanu, D. V.(1996) J. Biol. Chem. 271, 10508–10514

14. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith,J. A., and Struhl, K. (1994) Current Protocols in Molecular Biology, WileyInterscience, New York

15. Kumar, V., and Chambon, P. (1988) Cell 55, 145–15616. Schindler, C., Fu, X. Y., Improta, T., Aebersold, R., and Darnell, J. E., Jr.

(1992) Proc. Natl. Acad. Sci. U. S. A. 89, 7836–783917. Weiner, M. P., Costa, G. L., Schoettlin, W., Cline, J., Mathur, E., and Bauer, J.

C. (1994) Gene 151, 119–12318. Kanno, Y., Kozak, C. A., Schindler, C., Driggers, P. H., Ennist, D. L., Gleason,

S. L., Darnell, J. E., Jr., and Ozato, K. (1993) Mol. Cell. Biol. 13, 3951–396319. Umesono, K., Murakami, K. K., Thompson, C. C., and Evans, R. M. (1991) Cell

65, 1255–126620. Aguet, M., Gresser, I., Hovanessian, A. G., Bandu, M. T., Blanchard, B., and

Blangy, D. (1981) Virology 114, 585–58821. Yan, R., Qureshi, S., Zhong, Z., Wen, Z., and Darnell, J. E., Jr. (1995) Nucleic

Acids Res. 23, 459–46322. Carter, M. E., Gulick, T., Moore, D. D., and Kelly, D. P. (1994) Mol. Cell. Biol.

14, 4360–437223. Richard, S., and Zingg, H. H. (1991) J. Biol. Chem. 266, 21428–2143324. Vasios, G., Mader, S., Gold, J. D., Leid, M., Lutz, Y., Gaub, M. P., Chambon, P.,

and Gudas, L. (1991) EMBO J. 10, 1149–115825. Lipkin, S. M., Nelson, C. A., Glass, C. K., and Rosenfeld, M. G. (1992) Proc.

Natl. Acad. Sci. U. S. A. 89, 1209–121326. Kato, S., Sasaki, H., Suzawa, M., Masushige, S., Tora, L., Chambon, P., and

Gronemeyer, H. (1995) Mol. Cell. Biol. 15, 5858–586727. Palmer, C. N. A., Hsu, M.-H., Griffin, K. J., and Johnson, E. F. (1995) J. Biol.

Chem. 270, 16114–1612128. Naar, A. M., Boutin, J. M., Lipkin, S. M., Yu, V. C., Holloway, J. M., Glass, C.

K., and Rosenfeld, M. G. (1991) Cell 65, 1267–127929. Nagpal, S., Saunders, M., Kastner, P., Durand, B., Nakshatri, H., and

Chambon, P. (1992) Cell 70, 1007–101930. Williams, G. R., Harney, J. W., Moore, D. D., Larsen, P. R., and Brent, G. A.

(1992) Mol. Endocrinol. 6, 1527–153731. Mader, S., Leroy, P., Chen, J.-Y., and Chambon, P. (1993) J. Biol. Chem. 268,

591–60032. Kurokawa, R., Yu, V. C., Naar, A., Kyakumoto, S., Han, Z., Silverman, S.,

Rosenfeld, M. G., and Glass, C. K. (1993) Genes & Dev. 7, 1423–143533. Mader, S., Chen, J. Y., Chen, Z., White, J., Chambon, P., and Gronemeyer, H.

(1993) EMBO J. 12, 5029–504134. Perlmann, T., Rangarajan, P. N., Umesono, K., and Evans, R. M. (1993) Genes

& Dev. 7, 1411–142235. Zechel, C., Shen, X. Q., Chambon, P., and Gronemeyer, H. (1994) EMBO J. 13,

1414–142436. de The, H., Vivanco-Ruiz, M. M., Tiollais, P., Stunnenberg, H., and Dejean, A.

(1990) Nature 343, 177–18037. Mendelsohn, C., Larkin, S., Mark, M., LeMeur, M., Clifford, J., Zelent, A., and

Chambon, P. (1994) Mech. Dev. 45, 227–24138. Bromberg, J. F., Horvath, C. M., Wen, Z., Schreiber, R. D., and Darnell, J. E.

(1996) Proc. Natl. Acad. Sci. U. S. A. 93, 7673–767839. Chin, Y. E., Kitagawa, M., Su, W.-C. S., You, Z.-H., Iwamoto, Y., and Fu, X.-Y.

(1996) Science 272, 719–722

FIG. 8. Induction of STAT-1 promoter by RARbzRXRa combi-nation. A, reporter gene expression. Cell transfection, RA-treatment,and luciferase assays were performed as described in Fig. 4. COS-7 cellswere electroporated with VKL-7 reporter (5 mg) and the indicatedeffector plasmids (0.5 mg). The total amount of DNA transfected wasnormalized with pSG5. B, binding of RAR and RXR proteins expressedin COS-7 cells to STAT1-RARE. Cells were electroporated with 25 mg ofthe indicated reporter plasmid, and whole cell extracts were preparedafter 48 h (15). Extracts (15 mg) were assayed for DNA binding in EMSAwith RW probe. Lane 5, 7.5 mg of extract from each preparation wasincubated with the probe. Position of the specific complex is indicatedwith an arrow.

Retinoid Induction of STAT1 Gene9748